RESISTANCE PREDICTION USING ARTIFICIAL NEURAL NETWORKS FOR PRELIMINARY TRI-SWACH DESIGN

Abstract

Due to the novel hull form design, at present no standard series or full-scale data is publicly available to predict Tri- SWACH resistance during the preliminary ship design process. This work investigates the viability of using an Artificial Neural Network (ANN) to quickly predict total resistance for preliminary Tri-SWACH design.

An ANN was trained using total resistance experimental data obtained from model tests, which varied side hull arrangements. The results highlight strong correlation for model resistance prediction. A Tri-SWACH case study was then developed which had side hull geometric properties different to any previously used to train the ANN. The results, validated against CFD predictions, mimicked the resistance pattern generated by other model experimental data, providing confidence in the ANN’s ability to function as a resistance prediction tool.

This work demonstrates the viability of ANN to assess Tri-SWACH resistance as part of a preliminary design process. These results suggest that ANNs can be effective tools for assessing performance given relevant training data.